Closed songyuc closed 4 years ago
It is a standard procedure for anchor-free detectors such as CornerNet/ExtremeNet/CenterNet. These models use Hourglass backbone to generate high-resolution output featuremaps, use a higher input resolution will result in much higher computation cost.
Hi, @zzzxxxttt , thanks for your guide and answer! I think keeping the aspect ratio would be a better choice and seems more reasonable.
The preprocessing keeps the aspect ratio of origainal images, it uses zero padding to make them have the same resolution.
@zzzxxxttt , let me see. More specifically, do you mean the cv2.warpAffine() function will do the zero padding for the absent part of the image?
Yes
Hi guys, After reading the code of
coco.py
, I find that it seems the affine transformation in thecoco.py
forces the image into the size of 512x512, asI am not sure whether this is a general trick to resize the images of different sizes.
Any idea or answer will be appreciated!